Empiricist Sour Grapes

Comments on Cowie's What's Within?: Nativism Reconsidered

by Lawrence Kaye

Philosophy Department
University of Massachusetts at Boston


Cowie's book contains thorough and accurate expositions of Fodor's radical concept nativism and of Chomsky's linguistic nativism; it is particularly useful to see the course of development of Fodor's views of what concepts are and how we attain them. Cowie also does a nice job of presenting the theories and evidence concerning language acquisition. However, I do not find similar merits in her critical evaluations of nativism. In particular, her three main conclusions, viz., that nativism has two main faces, one of them anti-naturalistic, that there is an empiricist alternative to Fodor's concept nativism, and that we ought to pursue some sort of moderate nativist or empiricist account instead of Chomsky's theory of an innate universal grammar, are neither well developed nor convincingly supported. I will discuss each in turn.

1. What is Nativism?

Nativism, one might think, is a doctrine about the origin of mental states. While the empiricist claims that most or all mental states derive from the influence of the environment or experience, the nativist contends that these mental states are determined by genetic make up or are built into the mind/brain. However, Cowie argues, the title of her book not withstanding, that this way of characterizing nativism is inappropriate. Her main reason for rejecting this portrayal is that when we examine the actual views of nativists the inner origin view does not allow us to reconstruct a genuine disagreement with empiricism but instead shows the two parties to be merely talking past one another, emphasizing different aspects of the acquisition process. Cowie supports this contention with an examination of the Locke-Leibniz exchange.

After arguing against this conception of nativism, she proceeds to develop two alternative themes or "faces" that she maintains nativists slide between. The first is the hypothesis of special(ized) faculties to explain a given area of knowledge, a hypothesis based on so-called poverty of the stimulus arguments. Cowie attributes the special faculties view to Plato, Leibniz and Chomsky. The second face is wholly negative--a denial that naturalistic explanation of acquisition is at all possible. She finds this theme in Descartes, Leibniz and Fodor.

I think this conceptualization of nativism is mistaken; it is a caricature from someone ideologically mired in empiricism. As we will see, there is a reasonable way of developing the seemingly obvious inner-outer nativist-empiricist dispute that shows nativism to be a coherent albeit often underdeveloped view that is not inherently anti-naturalistic.

Let me begin with the question of what a good characterization of nativism should involve. Early in her discussion, after noting (some of) the variety of mental states that might be held to be innate and after skimming the various nativist metaphors about innateness, viz., "inborn", "imprinted" and "dispositional", Cowie complains that "the doctrine of innate ideas, which at first appeared to be a univocal position, has degenerated into a motley disjunction" (p. 5) Her desiderata is an exposition that would allow us to see various nativists as "saying--even roughly--the same thing" (ibid.)

Cowie seems to think that it is obvious that a correct account of nativism should yield a roughly equivalent view across various nativists. But it is not clear why this should be so. Take a typical philosophical school of thought such as realism, idealism, foundationalism or functionalism. Once we get past the broad strokes--the slogans and the metaphors--we find that each individual philosopher has developed the doctrine in a unique and idiosyncratic way. And that is what makes the study of philosophy so challenging and rewarding. But this is not to deny that there is something informative behind the names. What all (or most) adherents of an "ism" agree to is a general conception, e.g., mind-independent existence of certain entities or a purely relational account of mental states. But this broad conception does not serve to provide actual explanations, e.g., "how do you explain qualia?" So each thinker develops the broad notion in a more or less unique way in an attempt to produce a successfully explanatory view. Typically, while adherents of a specific "ism" share many specific doctrines, each does not hold them all. So what we have is not exactly a "motley disjunction", but rather a cluster of doctrines linked by family resemblance (which, N.B., tends to be generally true of non-natural kinds) that are in each case meant to fill out the general conception.

Not surprisingly, nativism consists of a cluster of related doctrines too. But before we examine them, consider that while Cowie contrasts nativism with empiricism, often unfavorably, she never explains what empiricism is. If we scope out a range of empiricist views, we will find a similar "motley disjunction". Thus, what theory do Locke, Berkeley, Hume, Mill, Mach, Quine, Skinner, Putnam and the connectionists all share? Not associationism, conditioning, conceptual reductionism, mentalism or even belief in the external world!

There thus seems to be no simple essence to empiricism either. But even though there is no simple doctrinal disagreement between empiricists and nativists, it is apparent that there are a variety of substantial specific disagreements between many nativists and empiricists. That is, I am suggesting that these are contrasting views that can only be properly understood by a simultaneous exposition.

Here, then, are the main elements of the cluster of issues that, I claim, respectively delineate empiricism and nativism.

 

    1. Empiricists postulate minimal mental structure concerning development whereas nativists postulate substantial structure. For instance, Hume postulates a set of simple ideas (impressions) and a mechanism of association. By contrast, contemporary modularity theorists postulate a host of perceptual units devoted to different tasks such as face recognition or grammatical parsing, linked together in complex ways.
    2. Empiricists postulate maximal environmental influence whereas nativists postulate minimal environmental influence. For example, typical conditioning theorists see learning as involving repeated trials that gradually shape behavior, while Fodor's triggering account of concept activation requires only a single encounter with an exemplar of the concept.
    3. Empiricists postulate a universal acquisition mechanism for all areas of knowledge whereas nativists postulate "domain specific" mechanisms. Here, an obvious contrast is universal associative learning versus Chomsky's theory of a language faculty.
    4. Empiricists postulate a uniform, undifferentiated knowledge structure whereas nativists postulate isolated units of knowledge--consider Quine's holistic web of belief versus the encapsulated knowledge of modularity theories.
    5. Empiricists postulate mental states that are plastic, and, in particular, are malleable to environmental influences whereas nativists postulate rigid mental states that remain constant through a variety of environmental influences. Here we may contrast Quine's view that our knowledge of our language grows and changes constantly throughout our lives with the Chomskyan view that our language faculty is programmed to acquire a grammar by certain age and that this grammar remains constant for the rest of the individual's life.
    6. Empiricists postulate a base of primitive building blocks that are the basis for all other mental states so that the latter can be analyzed or reduced to combinations of the former whereas nativists postulate a much larger set of mental states that do not have this compositional structure, as in Locke's versus Fodor's view of concepts.
    7. Empiricists postulate mental states that may vary from person to person whereas nativists postulate mental states that are universal. Here we may compare Locke's view that our ideas of substances (i.e., natural kinds) are arbitrary and variable with Descartes' view that we all possess certain innate ideas such as the idea of God.
    8. Empiricists postulate mental states (or behaviors) that arise gradually whereas nativists postulate mental states that arise quickly, perhaps instantaneously--consider Skinner's view of the gradual shaping of operant behaviors versus a one-encounter triggering view of concept attainment.
    9. And finally, empiricists postulate mental states that are acquired in the course of maturation whereas nativists postulate mental states that are present at or near birth. Here we may contrast any standard associationistic learning theory with certain developmental psychologists theories that depth perception and perhaps even object recognition are present within the first few months after birth.

Empiricists are thus theorists who develop the general conception of knowledge coming from the environment or experience, and their views typically involve the empiricist versions of (1-9) while nativists are theorists who develop the idea that knowledge comes from genes or the mind's built-in structure, and their views typically involve the nativist versions of (1-9). But the complexity does not end here, for one need not hold a view either in its extreme or its entirety. For instance, Skinnerians acknowledged the innateness of reflexes and Fodor is more or less a Quinean empiricist in regard to belief fixation.

Now, within this complex cluster of views, there is room for much of the sort of mere point of emphasis disagreement that Cowie points to. But it should also be clear that there is the potential for real disagreement on each and all of (1-9); empiricists typically find themselves on one side of each of each issues and nativists on the other.

Cowie seems to think that there has been a "great historical debate" between empiricists and nativists and we need an account that demonstrates the genuine conflict. However, it is simply not apparent that there ever has been such a debate. The most prominent exchange, namely Locke-Leibniz, is a notorious case of two thinkers talking past one another. It is not sound practice to reject an exposition of empiricism and nativism, as Cowie does with the external-internal contrast, because the exposition fails to show a genuine conflict between these two. And even if there has never been a genuine debate with both sides engaging the other's claims, it does not follow that there are not two broadly opposed camps here.

Let me turn to Cowie's attempt to make sense of what is really at the heart of nativism, viz., the two faces of special purpose faculties and non-naturalism. I agree that the idea of special purpose faculties is one way that a nativist view is sometimes developed but her presentation of this as one of only two themes of nativism simply fails to acknowledge the many other aspects of nativism and points of difference with empiricism that have been noted above. As such, her view fails to provide a helpful exposition of many nativist views that did not develop faculty psychologies (or did not emphasize this part of their view), such as the Cambridge Platonists, but who nonetheless had clear disagreements with empiricists on other matters. And her development of this issue reads too much of Chomsky's view back into historical nativists and thus under credits Chomsky's brilliant development of this aspect of nativism. It is one thing to propose that we have, say, a special faculty for knowledge of language, it is quite another to develop a plausible and well-supported theory of the knowledge that such a faculty is thought to have, a theory that moreover can be applied towards developing models of how acquisition proceeds.

Further, it is important to see that the application of the poverty of the stimulus argument is not limited to the support of domain specific faculties. For instance, it can be argued that there is not enough uniformity in human environments to explain why we all invariably possess the same instance of a type of knowledge, or it can be argued that a given empiricist acquisition hypothesis is unworkable since the environment provides insufficient training and feedback opportunities, or insufficient opportunities for rapid attainment of some type of knowledge.

While Cowie's first face of nativism is merely incomplete, the second face is wholly contrived. The non-naturalism of Leibniz and Descartes is typical of non-materialists. Contemporary nativists, e.g., Chomsky or modularity theorists, seem to be as "naturalistic" as anyone else, whatever that amounts to, exactly. It is simply fallacious to generalize non-naturalism to all of nativism.

And, indeed, it is not apparent that Descartes was as much of a non-naturalist as Cowie seems to think. Consider this alternative take on the "impossibility" argument that Cowie sites as evidence for non-naturalism about concept acquisition. Descartes' statement that "in no case do [the senses] exhibit to us the ideas of things" (quoted by Cowie, p. 52) can be understood as the claim that it is false that the senses transmit ideas to the mind. Rather, as any dualist should agree, perception must be a matter of causing, i.e., activating, certain states of the non-physical substance. That is, there is no concept learning that is a matter of the world or senses literally transferring ideas into the mind. Note that this is something that Locke agrees with as well (cf. Essay 2.1.2-3, a passage Cowie cites but does not connect up with Descartes' argument). But this is not to deny the empiricist view that it is possible to combine ideas as the result of experience. Rather, as Fodor (1981) clearly explains, the difference between a nativist and empiricist ideationist is in the size of the innate idea base--very small (e.g., "simple ideas") for the empiricist, large for the nativist. So this argument may be interpreted in a way that in not at all non-naturalistic.

Even thought the non-naturalist face is an inappropriate caricature of nativism, we may grant Cowie that, historically, there have been few detailed developments of nativism, and she is thus right to complain that nativism often seems like no more than a denial of empiricism. Indeed, Chomsky's view may be the only success story thus far for the nativist in terms of postulating a plausible alternative explanation of knowledge attainment, as opposed to mere hand waving metaphors. But the nativist suffers from being committed to genes or the structure of the mind as the cause of mental states, which requires an detailed understanding of one or maybe both in order to formulate a specific nativist doctrine. We may, however, be on the brink of genuine genetic causal explanation of aspects of the mind/brain. If so, then this century may bring the first full-blown nativist genetic origin theories, providing nativists with new ways of developing and explaining the general conceptions (1-9) that divide them from empiricism.

2. Concepts, Meaning and Reference

Cowie's discussion of Fodor's radical concept nativism contains two main undertakings. The first is an exposition and critique of Fodor's views of concepts and acquisition, viz., his (so-called) Standard Argument against empiricist acquisitions theories, his circa (1981) triggering view of innate concepts and his current (1998) view of concepts and their attainment. While, as I have discussed above, I feel that it is inappropriate to generalize Fodor's present anti-naturalist tendencies to the whole of nativism, I otherwise find this to be a useful and forceful evaluation of Fodor's position. (I provide a response to the Standard Argument in Kaye (1993); and, for more on all these matters, see the LOT entry.) However, Cowie's other project does not contain the same virtues. After showing us what is problematic with Fodor's account, she proceeds to sketch an account of concept acquisition in an effort to meet Fodor's Standard Argument, and claims to have shown that an empiricist view of concept attainment is both possible and plausible.

Unfortunately, confusion reigns in the latter investigation. Once we have sorted out Cowie's claims about concepts, meaning and acquisition, we will see that her suggestions do not actually solve any of the standard problems that concept acquisition theories face, and thus do not meet the challenge of Fodor's Standard Argument.

Central to Cowie's development of an empiricist alternative to Fodor's nativism is her understanding of Fodor's main anti-empiricist argument--Fodor (1998) labels this the Standard Argument against empiricist concept acquisition views. Cowie's summation of it runs as follows:

    1. Only internally structured concepts could be learned.

    2. Most concepts are indefinable.

    So, 3. Most concepts are not internally structured.

    So, 4. Most concepts could not be learned. (p. 76)

She contends that (3) does not follow from (1) and (2), since concepts may have non-definitional structure. And, she claims, many concepts do have another form of structure, viz., they consist of prototypes, which, she assumes, are learnable.

But both her reconstruction of the argument and her response to it reveal a fundamental misunderstanding of the issue here. As Cowie notes, concepts are not, strictly speaking, what gets defined. Words get defined and their definitions are thought to correspond to the structure of concepts. What's absolutely key here is that were are talking about the supposed semantic structure of concepts. The standard empiricist view, straight out of Locke and Hume, is that learning concepts is a matter of associating or connecting already known concepts, e.g., from one's stock of unstructured, innate primitive concepts. Moreover, the idea of concepts being built out of simpler ones seems to require that the more complex concepts reduce to their constituents, so assuming we have access to our concepts, we should be able to provide analytic, reductive definitions of the words that correspond to learned concepts. But we can't. Moreover, the very idea of reductive analytic definitions has been notorious challenged by Quine, and it is not apparent that anyone has produced a satisfactory response to the challenge. (Indeed, Fodor's concept nativism is perhaps best understood as a logical development of Quine's view that there is no meaning reductionism. Instead of opting for eliminativism about meaning, Fodor abandons empiricism for concept nativism.)

So a better reconstruction of Fodor's argument is:

    1. Only semantically structured, i.e., semantically composite, concepts could be learned.

    2. Most (single) concept terms don't have reductive definitions.

    So, 3. Most concepts are not semantically composite.

    So, 4. Most concepts could not be learned.

This makes it clear that a response to the argument requires more than just sketching apparent conceptual structure; one must present a case for composite semantic structure. And if twentieth century American philosophy of language is any indication, that is a very tall order.

Cowie never makes it clear if she is indeed proposing that prototypes constitute the semantic structure of concepts. I suspect that this reflects an apparent misunderstanding that has arisen generally between philosophers and psychologists. A theory of concept acquisition, or concept anything, for that matter, requires an understanding of what concepts are. Fodor (1998) anchors himself clearly in the Fregean (and Lockean) tradition by conceiving of concepts as mental particulars that are the units (or elements) of content-bearing thoughts and that are (or represent) categories. Philosophers, of course, tend to focus on the semantic issues. Psychologists, on the other hand, have tended to focus exclusively on the category issue; virtually all of the work on prototypes conceives concepts as knowledge structures, i.e., our knowledge of categories.

Now, while it's very plausible to think that we have prototype representations and they play important role in how we categorize the world, the claim that these prototypes are semantically constitutive of concepts is at best highly problematic. Suppose (Cowie's example) that we have a prototype for dogs involving the features of a golden retriever. If this is a semantic part of our concept DOG, then it should be an analytic truth that some dogs have golden fur or tails precisely the shape of a retriever's tail. But, of course, those aren't candidate analytic truths. Nor do prototypes provide the right truth conditions, because while poodles are much less similar (in my estimation) to golden retrievers than are labrador retrievers, all three are equally, 100% dogs. That is, if resemblance to the golden retriever prototype determined what counted as a dog, then poodles should be less of dogs than retrievers, but they aren't. And, if that weren't enough, as Fodor argues and Cowies discusses, prototypes don't compose--a representation of a golden golden retriever clearly does not participate in our comprehension of the phrase "small European dogs", and, indeed, we should expect a different prototype for that apparently semantically complex phrase. So, to repeat, it requires a lot of argument to make it even a bit plausible that prototypes are semantically constitutive of concepts, especially for natural kinds.

It is not at all clear whether Cowie does indeed want to claim that prototypes are semantically constitutive of concepts. She seems to think that her view entails semantic verificationism, and that that's a bad thing (pp. 140-1). (Incidentally, a major reason for avoiding semantic verification ism is that it appears to be false--people can possess concepts, like JUSTICE, while being unclear about what exactly counts as an instance.) But verificationism does not follow if the prototypes merely fix the reference of terms, for then the meaning of, e.g., 'dog' is the natural kind that the term (or mental representation) is causally, counterfactually connected to. On this sort of view, let us call it radical semantic externalism, I as a speaker who used the dog prototype to connect up with the natural kind dog may not have knowledge of the full extension that I'm connected with.

On the other hand, when Cowie is trying to dodge the bullet that protoypes don't satisfy compositionality, she writes:

    [W]hat compositionality really requires is that concepts contribute their references to complexes in which they occur. For what really matters to the explanation of productivity and systematicity is that the references of complexes depend in regular ways on the references of their constituent concepts. (P. 142, emphasis in original)

She goes on to distinguish meanings in the "intuitive" sense as that which "one must 'grasp' when one has a concept." (Ibid.) These are what enable us to "lock" to referents. Cowie then throws herself in with those who would abandon the Fregean view that meaning determines reference, and suggests that concepts in the intuitive sense need not compose, since it is only reference that requires compositionality.

But Cowie can't have it both ways; moreover, neither way is any good. The options, then, are 1) prototypes are at least partially semantically constitutive of concepts, in other words, the truth conditions of beliefs depend on the prototypes that make up the concepts that in turn make up the beliefs. 2) prototypes are used to fix references but are not part of the truth conditions. We have already noted that problems associated with (1), viz., prototypes do not seems to be yield appropriate semantic constitutive properties, like analytic truths about the prototypical features, prototypes don't compose, and they don't yield appropriate extensions for natural kinds. Cowie provides no help for these enormous difficulties.

What if Cowie were to, then, fully embrace (2)? First, consider what sort of response this is to Fodor, who argues that concepts, conceived as semantic units, for the most part cannot be learned. Cowie's reply in effect would be that concepts are not actually semantic units. So this is not empiricism, but eliminativism--she has not shown how Fregean concepts can be learned, but instead has asserted that concepts are not Fregean.

But can she, nonetheless, defend empiricism separately for reference and for concepts in the intuitive sense? Consider reference fixing. Cowie does not seem to be aware that the "qua problem" that she raises for Fodor's view applies equally to the view that she sketches. If I learn the reference of 'dog' by thinking "dogs are those kinds of things" while representing golden retrievers, why doesn't this connect 'dog' up with retrievers or pets or animals or prototypical dogs rather than with dogs? If we try to flesh out the story about what's in the head, in effect we're just expanding the story about concepts, and we run into the problems with (1) noted above. If for instance, we require that the learner think dogs are those kinds of ANIMALS, then we imply, problematically, that it's analytic that dogs are animals. And, assuming we're defending empiricism, we'll need a prior and independent acquisition story for ANIMAL, which is a dubious requirement since children appear to learn "middle-level" concepts like DOG prior to learning more abstract or concrete concepts. On the other hand, it may just be a brute-causal fact that such a ritual connects me up the property of doghood rather than some other property. But then there's no reason to call this an empiricst view. Indeed, then, there's little to distinguish this from Fodor's "anti-naturalistic" view.

It's thus hard to see how the externalist can be an empiricist about meaning, since as we considered, two of the typical elements of empiricism are compositional mental representations and trial and error approximation towards a goal, but the externalist act of reference fixing, and for that matter of reference borrowing, reflects neither empiricist element. Nor does the view imply or suggest any of the other typical features of empiricism.

Can Cowie at least claim hope for an empiricist story of concepts in the intuitive sense? A major problem that she overlooks is that the tradition of compositionality for meaning is based on the need to explain, not Tarskian truth conditions, but the ability to grasp or understand meanings. Thus, a standard argument for compositionality is that we're able to comprehend infinitely or indefinitely many sentences, but the obvious limits on the brain's cognitive capacity show that we must do it by grasping a base set of meanings that yield more complex sentences compositionally. Since this argument is based on understanding rather than on reference or truth, it applies to concepts in the intuitive sense. So, if as Cowie seems to admit, prototypes don't compose, then she hasn't given an empiricist explanation (or any explanation) of concepts in the intuitive sense either.

And while we are on the subject of radical semantic externalism, allow me to sneak in my internalist bias and point out the implausible implication that externalism has: our apparent experience of meanings is merely phenomenal. Any sense of grasp or understanding or the like we may have when encountering a sentence is no guarantee that we are actually connected up with the appropriate truth conditions, and indeed, concepts in the intuitive sense may be completely out of line with reference. If radical externalism is correct, then there is no reason to think that ordinary speakers have any demonstrative knowledge of the truth conditions of their words. On the scales of implausibility, I don't see that this weighs any lighter than Fodor's radical concept nativism.

So, to sum up, Cowie's attempt to provide an empiricist response to Fodor's impossibility argument is completely implausible since it doesn't tackle, let alone solve, well-known problems with the semantics of prototypes, is strictly speaking not an answer to Fodor, and does not really amount to empiricism anyway. Fodor's challenge is a really hard problem for empiricists.

3. Linguistics and the Philosophy of Science

In the third part of her book, Cowie evaluates Chomsky's view that humans possess an innate language acquisition device (LAD) containing a universal grammar (UG). She rejects the primary Chomskyan argument for this view, viz., the poverty of the stimulus argument, and finds much of the evidence typically cited by Chomskyan linguists to be problematic at best, yet concludes that the most likely hypothesis is that there is indeed some sort of special language faculty, though not necessarily one that possesses an innate universal grammar. She thinks she has shown that Chomsky's view is less well-supported than is usually assumed, and that weaker nativist and empiricist alternatives ought to be investigated.

What I find problematic in this part of her book is the whole framework for evaluation. Chomsky's views of language, both general and specific, constitute a scientific theory. An assessment of his outlook therefore presupposes a conception of what scientific theories are and how they can be evaluated. Cowie seems to hold an old-fashioned deductive nomonological view of science, since she looks for decisive evidence and arguments that would either support the LAD/UG view or discredit empiricist alternatives. However, in these enlightened post-Kuhnian times, we should be very wary of that sort of approach.

A revealing passage is this regard is where Cowie, in introducing the five core claims she finds in Chomsky's view, tells us that:

    Chomskyan nativism embraces a number of partially independent theses concerning language-learning and thus it is possible rationally to reject some elements of the Chomskyan position while accepting others. Indeed, I'll ultimately suggest, since Chomsky's "core" claims as a matter of fact possess different degrees of evidential support, picking and choosing among them is exactly the right thing to do (p. 154).

However, the last four decades of philosophy of science (especially Kuhn 1962) have taught us that scientific theories do not rest on isolated evidence, but rather depend on connections to related theories, background assumptions and specialized methods of investigation. That is, they form world views. And this is a very fitting conception of Chomsky's complex outlook.

While it may be rationally possible to pick and choose individual claims from a scientific outlook, and that is certainly something that us armchair philosophers are wont to do, it is quite another matter to insist that this is how science ought to operate, for, as a matter of fact, it doesn't.

What does make for good science, if not isolated support for isolated theories? To borrow a page from Lakatos (1978), it is research programs that form the units of success and failure. To put it simply, successful research programs survive--they beat out their competitors. And they do so not just by providing satisfying explanations and supporting evidence, but by producing periods of productive, developmental detail work on the theory--what Kuhn (1977) terms "normal science"--and occasional "revolutions" that result in broad revisions to the view and open up more and better detail work.

It is, I think, very easy to evaluate Chomsky's view of language as a scientific research program. It has generated and continues to generate lots of detailed research. In this regard it is undoubtedly the most successful linguistic theory we've ever had. And, thanks to Chomsky himself, it has not stagnated; there have been several major revolutions including the "parameters and principles" shift. (Indeed, Cowie fails to chronicle the most recent revolution--Chomsky (1995)).

Most of all, Chomsky's approach has come to completely dominate linguistics; rival research programs are few and far between, and in particular, there are no largely successful, long-running rival empiricist approaches. So, on a reasonable conception of the nature of science, we are forced to conclude that Chomsky's nativism has won the day, period, end of story.

Now, we, as philosophers would very much like to know what the dominant paradigm for language will be in 100 or 1000 years. But, to echo a point I heard Chomsky make years ago, the truth is, there really is no way predict this. That's what it means to have an empirical rather than an a priori discipline.

Cowie does, though, compare Chomsky's LAD/UG view with three rivals: the "Putnamian empiricist" who appeals to "general multipurpose learning strategies" for all of language-learning, the "enlightened empiricist" who accepts that there are special principles involving langauge-learning but thinks that they are learned as well and the "weak nativist" who believes that there are innate, language-specific constraints on language learning but rejects Chomsky's view of UG as too rich a description of what is innate.

These are, however, merely imagined theories. (And this is an odd quirk that philosophers have--we'll argue about and even believe in merely imaginary theories, e.g. the identity "theory" or the new "theory" of reference.) In particular, the only general learning theory we've ever had, viz., associationism, has been a colossal failure, through several revivals across several centuries. How are we to imagine the Putnamiam empiricist for instance--as a successful associationist? Why should we believe that is possible? Or as having developed a successful alternative to associationism as a general learning theory? What would such a theory look like? How can we hope to assess it vis-a-vis the evidence or compare it with Chomsky's view unless we have a pretty clear idea what it says about the mechanisms and procedures of learning?

To see why this matters, note again that the poverty of the stimulus argument is a very strong argument against the typical and traditional rival empiricist approach, viz., associationism. The latter view conceives of gradual shaping through trial and error learning with lots of feedback. The facts that there are few errors and little feedback thus count decisively against a standard associationistic approach, in particular the behavioristic conditioning views that dominated the previous generation of linguistics. And, incidentally, the same facts are proving very problematic for contemporary connectionist approaches that also involve gradual shaping techniques and initially postulated massive feedback. Again, without an actual theory there can be no assessment.

Perhaps, though, Cowie simply hopes to inspire someone to develop a new and more successful empiricist outlook. That's fair enough, but we might wonder, why bother? Cowie seems to assume throughout that empiricism is a sort of plausible default view, i.e., that there's independent reason to think that there must ultimately be a successful empiricist theory. For instance, she says that "learning tout court is the main unsolved mystery of modern cognitive science (p. 306)." But it's important to see that the whole idea of there being "learning tout court" is an open question. And I, for one, don't think there is any such thing.

Consider, for instance, Cowie's remarks that:

    language displays certain marks that are suggestive of its having a nativist basis: the existence of language universals, the dissociation between children's acquisition of language and their acquisition of other skills of comparable complexity, phenomena like creolization. But there are other abilities that we have that do not display these sorts of characteristics and that seem more plausibly viewed as products of a powerful general capacity we humans have to learn from experience. Knowing what curries are; knowing one's multiplication tables and how to use them; recognizing a Mozart symphony or a David Bowie lyric; distinguishing an electron's track in a cloud chamber from that of an alpha particle; or picking the Chanel from among the Diors; writing a philosophy paper; or driving a car; or singing the National Anthem--these are abilities which it seems, to say the least, far-fetched to postulate special faculties (p. 305).

Au contraire. While it's unreasonable to think that there are ultra specific faculties, e.g., a curry faculty, a good case can be made that each of these abilities either involves knowledge from a specialized faculty (or module) or otherwise involves some innate principles. In turn:

    Curry A major part of knowing curries involves knowing their taste, and It's plausible to think that perception involves lots of specialized units (modules) with innate principles and processes and certainly that includes taste.

    Multiplication tables It's plausible to postulate a specialized mathematical faculty. Moreover, there has been a lot of research on the acquisition of basic arithmetical skills; they appear to be specialized and universal and are thus good candidates for innateness. (Gelman and Gallistel, 1986)

    Mozart A specialized music faculty has been postulated as has a generative theory of tonal music (Lerdahl and Jackendoff 1983).

    Lyrics It's reasonable to think that memorizing sentences and phrases involves a separate, specialized ability distinct from remembering the content of what is said. It's also likely that appreciation of the cadence of a song lyric (well, from good songwriters like Dylan or Willie Dixon anyway) involves the musical faculty.

    Track in a cloud chamber Specialized perceptual shape-detection modules should be at play here.

    Perfume A major part of this ability will rely on perceptual odor modules.

    Writing a Philosophy Paper Those of us who teach philosophy know that this is a highly specialized skill, and we see how a few take to it readily while most struggle terribly. It is also apparent that a major part of being able to do philosophy well involves having been born with an unusually good critical thinking faculty.

    Driving This involves coordination of motor behaviors and commonsense has it that the degree of coordination that one possesses is largely innate.

    Singing the National Anthem Again, a music faculty. This is also an interesting example because it requires carrying a melody, which some people find easy and others impossible, as well as having a fairly large vocal range, which notoriously does not vary much even with lots of training.

Moreover, these are very interesting examples because it's clear that someone could be very good at one of these abilities--pick it up easily--and yet be very, very bad at another, finding it impossible no matter how hard they try. If there is a general learning mechanism behind all of these abilities, then why should this be? If empiricism is generally true, why shouldn't it be the case that anyone can learn to sing beautifully or be perfectly coordinated or a culinary expert or a great philosopher? Varying innate abilities, and lots of them, is the very, very obvious explanation here.

I suggest, then, that there really are no good arguments or evidence for empiricism. It's just several centuries of philosophical dogma that inclines us to think it has some enormous plausibility.

In any case, in science, the name of the game is: put up or shut up. As Cowie ultimately states, "it's unclear how we learn anything!"(p. 306) But that, it seems to me, is to admit that there are at present no actual, plausible empiricist theories, for language acquisition or anything else. So, at least for the moment, nativism is where it's at.

      References

Chomsky, Noam (1995) The Minimalist Program. Cambridge, MA: MIT Press.

Fodor, Jerry A. (1981) "The Present Status of the Innateness Controversy", in his Representations, Cambridge, MA: MIT Press.

Fodor, Jerry A. (1998) Concepts: Where Cognitive Science Went Wrong. New York: Oxford University Press.

Gelman, Rachel and C.R. Gallistel (1986) The Child's Understanding of Number. Cambridge, MA: Harvard University Press.

Kaye, Lawrence J. (1993) "Are Most of Our Concepts Innate?", Synthese, 95, 2, pp. 187-217.

Kuhn, Thomas S. (1962) The Structure of Scientific Revolutions. Chicago: University of Chicago Press.

Kuhn, Thomas S. (1977) The Essential Tension: Selected Studies in Scientific Tradition and Change. Chicago: University of Chicago Press.

Lakatos, Imre (1978) The Methodology of Scientific Research Programs: Philosophical Papers Volume 1, Worrall and Currie, eds. New York: Cambridge University Press.

Lerdahl, F. and R. Jackendoff (1983) A Generative Theory of Tonal Music. Cambridge, MA: MIT Press.

 top | back to symposium index | Cowie's reply